In partnership with
Qatar declared force majeure on LNG. Singapore powers data centres with that gas. Nobody connected the two.
On 18 March, Iranian strikes hit Qatar's Ras Laffan Industrial City. The complex produces a quarter of the world's Liquefied Natural Gas (LNG). QatarEnergy declared force majeure on its supply contracts. Global LNG supply fell 20% within weeks.
Singapore generates 95% of its electricity from imported gas. Sixty percent arrives as seaborne LNG. Data centres consume 7% of national electricity. In December 2025, the DC-CFA2 programme approved 200 MW of new data centre capacity. A 700 MW park on Jurong Island could expand total capacity by 50%.
Three months later, Qatar's force majeure hit. The government is now procuring emergency LNG cargoes from outside the Middle East. Every new megawatt of approved compute capacity increases the dependency on a supply chain that just fractured.
The LNG procurement team did not model data centre load. The data centre allocation team did not model LNG supply risk. A strike 6,000 kilometres away threatened compute uptime. Neither team knew the other's infrastructure existed.
This is the pattern. Three sectors share grid connections across four continents. None shares a schedule.
France hedged this blind spot before the crisis. Data4 signed 40 MW of nuclear. Nobody else followed.
Electricite de France (EDF) operates 57 reactors across 19 sites. The fleet generates 350 to 370 TWh annually. France's data centre market is worth £5.4bn ($6.9bn, €6.5bn) and growing at 9.7% per year.
In September 2025, Data4 signed a 12-year Nuclear Production Allocation Contract (NPAC) with EDF. Forty megawatts. 230 GWh per year. The first direct nuclear supply deal for a European data centre operator.
Six months later, Qatar's force majeure proved the thesis. Data4's compute runs on reactor physics, insulated from gas supply disruption. Singapore's compute runs on gas molecules that now transit a war zone. France's data centre market is worth £5.4bn ($6.9bn, €6.5bn) and growing at 9.7% per year. One operator hedged the cross-sector risk before the crisis had a name. The rest did not know they were exposed.
Stargate energises 1.2 GW this month. Gas plant costs rose 66% in two years. The grid did not plan for both.
The Stargate project in Abilene, Texas energises its first phase this month. The site targets 1.2 GW of AI compute. The broader programme spans five additional US sites, approaching 7 GW of planned capacity backed by £390bn ($500bn, €470bn).
Meanwhile, the cost to build gas power plants has surged 66% in two years, driven by data centre demand. Four hundred miles south, Cheniere's Sabine Pass operates at 30 million tonnes per annum (mtpa). Four new terminals under construction add 10 billion cubic feet per day of gas demand.
Both sectors compete for gas-fired generation on the Texas grid. Both drive transmission investment. Both face interconnection constraints. Neither sector's planning cycle references the other.
Crusoe expanded its Abilene campus to 1.2 GW, adding 4.5 GW of natural gas capacity across its portfolio. AI compute and LNG exports drive the same cost spiral. Nobody models the combined load.
Brazil generates 85% renewable electricity. Hyperscalers committed £3.5bn before anyone mapped the grid.
Brazil's electricity mix is approximately 85% renewable, dominated by hydropower at 62%. The data centre market is worth £2.7bn ($3.4bn, €3.2bn) and growing at 14.6% annually.
Microsoft committed £2.1bn ($2.7bn, €2.5bn) to expansion. Amazon invested £1.4bn ($1.8bn, €1.7bn). Twenty-three projects target completion by Q4 2026.
Brazil's cross-sector pattern differs. Hydropower provides cheap, low-carbon electricity. LNG plays a marginal balancing role. The data centre boom arrived before grid infrastructure caught up. Transmission constraints between the hydropower-rich north and the demand-heavy southeast remain the binding factor.
Capital arrived based on electricity costs. Nobody modelled the grid bottleneck between where power is generated and where compute needs it.
Which inter-sector proximity would matter most to your decisions?
683 facilities. Three sectors. The grid connection they share but cannot see.
Atlas tracks 683 facilities across nuclear, LNG, and data centre sectors spanning 31 countries and four continents. Each facility sits on a shared grid. Each draws from or feeds the same transmission infrastructure.
The pattern is consistent. In France, reactors and data centres share EDF's network. In Texas, AI factories and LNG terminals compete for gas-fired generation. In Singapore, LNG imports power data centres through two degrees of conversion. In Brazil, hydropower feeds both emerging sectors through constrained corridors.
No single-sector database captures this. The IAEA Power Reactor Information System (PRIS) tracks reactors. FERC tracks LNG terminals. Hyperscaler announcements track data centres. None maps the three together.
Atlas does. A nuclear plant's grid headroom matters to the data centre team scouting adjacent land. Only a multi-sector view shows it. When an LNG terminal's gas demand competes with a data centre's electricity supply, only a cross-sector lens reveals it. The engineer makes the call. Atlas shows what was invisible before the decision.
A 20-year LNG contract expires the same year as a 20-year PPA. The data centre lease outlives both.
LNG supply agreements run 15 to 20 years. Nuclear Power Purchase Agreements (PPAs) run 15 to 25 years. Data centre ground leases run 25 to 40 years. All three were signed in the same decade. They expire in overlapping windows.
A 20-year LNG agreement expires in 2040. The data centre it powers still needs electricity for 15 more years. A nuclear PPA expires the same year. The reactor runs for another 40 years. The data centre remains. The repricing risk cascades across all three sectors simultaneously. Qatar's force majeure proved what happens when the clocks intersect during a crisis. The LNG contract failed. The data centre still needed power. No contingency plan spanned both sectors.
Last week we examined how a 20-year PPA on a 60-year asset creates residual value the tenant never captures. The cross-sector version is worse. Three contract clocks. Three sectors. No shared model.
Atlas maps 683 facilities across all three. The cross-sector view is open at atlas.vistergy.com.
Quarterly facility-level exports launch next quarter for Intelligence Newsletter subscribers. The newsletter is the entry point.
Next week: Atlas shows what the satellite saw. Here is what it cannot.
The cross-sector pattern is visible from orbit. Satellite verification confirms construction status, thermal signatures, and operational changes. But every intelligence layer has limits.
Next week on the Vistergy Brief: what satellite monitoring reveals, and the five constraints it cannot overcome.
Full Atlas: atlas.vistergy.com
1,000+ Proven ChatGPT Prompts That Help You Work 10X Faster
ChatGPT is insanely powerful.
But most people waste 90% of its potential by using it like Google.
These 1,000+ proven ChatGPT prompts fix that and help you work 10X faster.
Sign up for Superhuman AI and get:
1,000+ ready-to-use prompts to solve problems in minutes instead of hours—tested & used by 1M+ professionals
Superhuman AI newsletter (3 min daily) so you keep learning new AI tools & tutorials to stay ahead in your career—the prompts are just the beginning